rooty/LibrePilot
1
0
Fork 0
mirror of https://bitbucket.org/librepilot/librepilot.git synced 2024-11-29 07:24:13 +01:00
Code Issues Releases Activity
4,816 Commits 5 Branches 8 Tags 266 MiB
47fac7e31c
Commit Graph

4816 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
James Cotton
47fac7e31c Move the throttle setting code for alt hold into the fast accel loop. 2012-02-18 18:16:15 -06:00
James Cotton
7d4582e5f0 Update the altitude KF to use the updated measurements appropriately and check 
in the relevant matlab code to check it and genererate it.  Produces quite
smooth traces.
 2012-02-16 05:33:47 -06:00
James Cotton
fbb0b1173a More work from Kenz to improve the matlab parser 2012-02-16 04:24:10 -06:00
James Cotton
c957172f8f In the case of swapping in userspace due to the buffer being blocked during 
ISR, make sure the latency is less than 50 us to avoid framesync errors.  Still
seeing a few nonsense packets on Overo side but fairly few.  Probably need to
add CRC to whole packet now and call it done.
 2012-02-14 11:06:16 -06:00
James Cotton
1930d092eb In the case where the overo scheduler cannot swap buffers (userspace loading 
it) then set a flag so as soon as the buffer is loaded then we schedule it.
 2012-02-14 11:00:29 -06:00
James Cotton
f04c58b0f4 Make the overo transaction occur based on the rising edge of the NSS line. 
This fixes most of the frame sync problems.  However, it's a bit hacky as
there is currently hardware information inside a module.
 2012-02-14 10:37:30 -06:00
James Cotton
fea4accb07 Patch from Kenz to suppress warnings about unknown object IDs until end of 
parsing
 2012-02-13 16:36:05 -06:00
James Cotton
81ceb9c17a Precompute CRC table in a global constant. Also if datasize for unknown object 
is > 256 ignore it since corrupt data.
 2012-02-13 16:27:44 -06:00
James Cotton
d018d93e88 Clean up previous commit a bit and make the sampling rate a variable (and adapt 
the sampling time accordingly).
 2012-02-13 01:49:54 -06:00
James Cotton
3acbfbe96c Convert the pressure sensor code to most of the calculations as int64 to retain 
precision.
 2012-02-13 01:22:48 -06:00
James Cotton
4717af1afd Swap telemetry port and gps ports back 2012-02-12 13:55:52 -06:00
James Cotton
94d76ea62f Bump overo submodule 2012-02-12 11:01:57 -06:00
James Cotton
f23f53dc4e Add settings for the altitude kalman filter settings and init the state when 
first measurement arrives
 2012-02-12 09:07:27 -06:00
James Cotton
34ca5777dd Kalman filter for altitude 2012-02-11 21:08:19 -06:00
James Cotton
d8f0494e57 Merge branch 'revolution' into revolution_overo 2012-02-11 20:02:20 -06:00
Alessio Morale
22b211167c Ported pios_iap.c to F4 
restored failed boot count check on pios_board.c for revo
 2012-02-11 13:49:32 +01:00
James Cotton
4d659fb2e9 Bump overo rev 2012-02-10 08:08:18 -06:00
James Cotton
6d6c7cac60 Added CRC calculation to the matlab log parsing 2012-02-10 08:07:51 -06:00
James Cotton
3026527801 Now the desired altitude is relative to when the switch was flipped so that the 
baro altitude module can use a smoothed altitude estimate for the starting
point.
 2012-02-08 23:02:29 -06:00
James Cotton
80d602ef7b Switch order of some commands to make sure it loads the correct hover throttle 2012-02-08 20:47:10 -06:00
James Cotton
ab41bae071 Separately smooth the error estimate from the derivative 2012-02-08 11:15:33 -06:00
James Cotton
60c951843d Use backtracking to limit integral in altitude hold and smooth the velocity 
error.  Initialize the throttle integral better.
 2012-02-08 10:39:31 -06:00
James Cotton
d5207c19f0 Fix for accidentally including inssetings back into build 2012-02-08 09:46:51 -06:00
James Cotton
c6b1d6b8df Rework the altitude hold code a bit 2012-02-08 09:42:10 -06:00
James Cotton
49a03a868b Merge branch 'altitudehold' into revolution_overo 
Conflicts:
	flight/OpenPilot/UAVObjects.inc
	flight/Project/OpenPilotOSX/OpenPilotOSX.xcodeproj/project.pbxproj
	flight/Revolution/Makefile
	flight/Revolution/ins.c
 2012-02-07 23:00:29 -06:00
James Cotton
6ec0f06c80 Back to 1024 byte packets 2012-02-07 13:02:11 -06:00
James Cotton
2c60447200 Switch back to 512 byte packets and fix bug when writing the header size that 
was smashing the crc
 2012-02-07 11:32:04 -06:00
James Cotton
221a0c5e53 Add the number of dropped updates to the overosync stats 2012-02-07 09:13:26 -06:00
James Cotton
a89d425b23 Reset the received rate for overo sync object 2012-02-07 09:01:58 -06:00
James Cotton
02bec750bd In the attitude loop need to check all the queues even if unused or the event 
system will have warnings
 2012-02-07 08:46:34 -06:00
James Cotton
3bb571a1a7 Delete overo stubs to move into submodule 2012-02-06 15:48:30 -06:00
James Cotton
ae01936808 Add the portTickTime into the OveroSync stream 2012-02-06 10:00:02 -06:00
James Cotton
dbcb499d7f Add object to monitor the overo link status. Current just records the bytes/s 
being sent and the connected status (based on sending more than 500 bytes per
second)
 2012-02-06 09:37:31 -06:00
James Cotton
5b2376dbee Remove the semaphore for buffer swapping and make this occur during the receive 
ISR.  This leads to a possible condition where a copy is taking place in
userspace and then a buffer swap interrupts it.  However, the copy should
always finish before the SPI DMA slave catches up to that place in memory at 10
MHz.

This fixes a common condition where the userspace task (low priority) was not
swapping buffers in time from user space and logs were corrupted.
 2012-02-06 08:21:24 -06:00
James Cotton
179e66b754 Refactor how new packets are scheduled to avoid trying to start a transaction 
from an ISR.  This was causing lock ups.  It's a bit higher latency probably
(trivially slow) but if we aren't keeping up with the Overo transactions then
using a proper DMA double buffer will be required.
 2012-02-04 20:11:31 -06:00
James Cotton
665801662d In cc3d update sensors add accelbias before scaling to allow more precision in 
levelinIn cc3d update sensors add accelbias before scaling to allow more
precision in levelingg
 2012-02-04 12:12:41 -06:00
James Cotton
4e190f99e2 Improved the revo overosync code to put multiple uavobjects into a packet 2012-02-04 12:06:03 -06:00
Stacey Sheldon
60a2af2e58 ef+dfu: Move EF image generation into separate makefile 
A separate makefile for EF (Entire Flash) targets allows
the removal of board-specific hard-coded addresses.  These
values can now be included from the board-info.mk files
for each board.
 2012-02-04 12:03:42 -05:00
Stacey Sheldon
4fafbb13ed makefile: Add EF (Entire Flash) targets for all boards 
These new build targets will construct a flash image that
includes both a bootloader as well as a firmware image all
in one.  Examples:
  make ef_coptercontrol
  make ef_revolution

This also adds support for a new target for writing this EF
image to boards that support a DFU interface (only F4-based
boards).  This can be done by running:
  make dfuutil_install
  # Boot board with SBL asserted to activate STM32 DFU ROM
  # Attach USB cable
  make ef_revolution_dfu

Bootloader images are also now padded out to occupy the full
BL flash region.
 2012-02-03 00:59:36 -05:00
Stacey Sheldon
ce544cf4ab gdb: remove cortex_m3 remnants from revo gdb init file 2012-02-03 00:59:31 -05:00
Stacey Sheldon
99f5f20cb2 openocd: add support for stlink on stm32f4 targets 2012-02-03 00:51:31 -05:00
James Cotton
18a64e261f Going a bit overkill with this - deinit who SPI block between transactions. 2012-02-02 12:32:57 -06:00
James Cotton
c26154f303 Switch back to 256 byte packets and 0xFF pad the end 2012-02-02 11:40:49 -06:00
James Cotton
03cfb1453a DeInit SPI DMA between transactions. Somehow if this doesn't happen the next 
transaction doesn't start at the beginning again.  Something is still not quite
right because first few bytes aren't getting across.  Need logic analyzer.
 2012-02-02 11:37:35 -06:00
James Cotton
77a6abb341 For PIOS_SPI F4 do not enable the TX irq. Also do not wait for the 
transmission to finish either.
 2012-02-02 11:24:14 -06:00
James Cotton
8b1137c6a0 Align the tx and rx SPI buffers from Overosync 2012-02-02 11:23:40 -06:00
James Cotton
0a71a48c57 Sending all FF finally again 2012-02-02 10:36:09 -06:00
James Cotton
80b839d145 The SPI3 TX DMA channel was wrong 2012-02-02 10:13:09 -06:00
James Cotton
29237f97ae For SPI slave devices we need to handle the interrupt differently since 
transactions should already be completed.  Also reset the callback so in the
case of noise on the IRQ line it will not keep firing interrupts.

I suspect we should probably disable interrupts in this handler to prevent
refiring.
 2012-02-02 10:13:09 -06:00
James Cotton
88b9f20cda Fix nasty bug in the DMA stream configuration 2012-02-02 10:13:09 -06:00